摄像机自标定技术研究
文献类型：学位论文
| 作者 | 雷成 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2001-11-01 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 胡占义 |
| 关键词 | 六点算法 摄像机自标定 三维重建系统 基本矩阵自动估计 Kruppa方程 Six-point Algorithm Camera Self-calibration 3D-reconstruction System Fundamental Matrix Automatic Estimation Kruppa Equation |
| 其他题名 | Study on Technipues for Camera Self-Calibration |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 作为计算机视觉技术研究的重要研究方向之一，摄像机自标定和三维重建技术有 着很高的理论研究价值和广泛的应用前景。到目前为止，国内外在此领域的研究取得 了丰硕的研究成果。但总体来说，仍存在着很多的理论和技术问题需要解决。有鉴于 此，本论文在前人研究的基础上，对目前摄像机自标定和三维重建研究领域中存在的 一些主要问题进行了探索性的研究，所完成的工作概括如下主要有： （1）研究了利用基于“三图象几何”的六点算法自动估计基本矩阵以及剔除自动图象 匹配过程中所不可避免产生的误匹配的方法。该方法良好的鲁棒性和精确性在实 验以及三维重建系统实现中均得到了体现； （2）研究了传统基于Kruppa方程的摄像机自标定方法欠鲁棒性的问题，通过大量仿真 实验揭示了导致该问题的三个内在原因； （3）提出了一种新的基于Kruppa方程的二步式摄像机自标定方法，改变了求解Kruppa 方程的原有思路，大大提高了原有算法的鲁棒性和实用性； （4）提出了一种新的基于主动视觉的摄像机自标定算法，从理论上证明了通过控制摄 像机平台作4次平移运动（其中任意三次不共面）便可以线性地标定摄像机的内外参 数； （5）基于本论文在基本矩阵自动估计、摄像机自标定等方面取得的研究成果，实现了 一套基于三幅图象的自动三维重建系统。一方面验证了本论文所提各种方法的正 确性，另一方面也为今后的技术产品化工作建立了系统原型。 |
| 英文摘要 | As one of the most important research directions, camera self-calibration and 3D reconstruction have a high research value and a wide application perspective. So far, the researchers have obtained fruitful achievements in this research field. However, there remain some theoretic and technical problems to be solved satisfactorily. In this thesis, we propose a number of new algorithms and techniques for camera self-calibration and 3D reconstruction. The contributions of this thesis include: (a) We applied the six-point algorithm originated from "tri-ocular geometry" to the problem of automatic estimation of the fundamental matrix when three images are available. From the good results of the conducted experiments and the promising performance of a developed 3D-reconstruction system, its robustness and accuracy has been proved. (b) We investigated the instability problem of the traditional Kruppa equation based camera self-calibration techniques. By conducting extensive simulations, we reveal three intrinsic reasons incurring the instability. (c) We proposed a new two-step camera self-calibration technique by changing the original approach for solving the Kruppa equations. This new technique improves the robustness and practicability of the Kruppa equation based self-calibration techniques. (d) We proposed a new active-vision based camera self-calibration technique. We prove that the camera's intrinsic and extrinsic parameters can be calibrated linearly by controlling the camera platform to undergo four translations with any three of them not being coplanar. (e) We developed an automatic 3D-reconstruction system based on our proposed techniques for fundamental matrix automatic estimation and camera self-calibration. The promising performance of this system proves the validity of these techniques. On the other hand, this system also offers a prototype for the future commercialization of our 3D-reconstruction techniques. |
| 语种 | 中文 |
| 其他标识符 | 659 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/5723]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 雷成. 摄像机自标定技术研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2001. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。